Re: References and Modularity

On Tuesday, June 4, 2013 at 1:31 AM, Jonathan A Rees wrote:

>  
>  
> On Sat, Jun 1, 2013 at 4:04 AM, Anne van Kesteren <annevk@annevk.nl (mailto:annevk@annevk.nl)> wrote:
> > On Thu, May 30, 2013 at 11:22 PM, Larry Masinter <masinter@adobe.com (mailto:masinter@adobe.com)> wrote:
> > > Is there any part of www.w3.org/TR/qaframe-spec/ (http://www.w3.org/TR/qaframe-spec/) you disagree with?
> >  
> >  
> > I think it makes matters too complicated. Referencing e.g. BCP47 for
> > languages tags is fine. There's no need to reference a dated version.
> > Standards for (web) software require active maintenance so using
> > references that won't change over a decade is the wrong optimization.
>  
>  
> There are two distinct use cases. If you want to say that information about how such and such is done has a particular name (e.g. BCP47), lives in some place, or is maintained by some organization, then the undated reference is OK since you really are talking about an ongoing process. However, there is also the situation that you find in scholarship: you (the author) are making a claim, the truth of which depends on the content of some document, and your reputation depends on the validity of your claim. In this situation it is very important to do the dated reference, since the ways in which the target document might change are beyond your control, and your claim might become false if evaluated against an updated version. Maybe you usually want to provide both references: X1 is the version of X we consulted, and the current version of X can be found (or is called) such and such.
In spec land, we should discourage the above (as it's a serious limitation in the broken paper-based publishing model). The Web Platform is a version-less system; all Web platform specs contain bugs and need to be actively maintained (hence the BCP47 style/model is more appropriate - Web specs are not affected by the limitations of paper based publishing as digital specs can be updated with no limitation. It's inexcusable to not fix things in place in a digital document, where the cost of republishing is near 0). The only use case for citing dated references is for non-normative things - if someone is citing dated versions for normative specifications, then they are generally doing it wrong (or are working with a publishing system that is doing it wrong, as is the current case with things like Respec - but we are working on fixing that!).

If you want to compare and contrast, this is the "right way"™ to do referencing in Web specs, IMO:
http://www.w3.org/TR/html51/references.html#references

(I would personally exclude the Editor's too from the references, as those can be gotten by following the link to the appropriate spec)


--
Marcos Caceres

Received on Tuesday, 4 June 2013 09:49:51 UTC